Migrate curriculum rig import to city.toml [[rigs]] (gc 1.2.1)

[austinborn]

Summary

gc 1.2.1 rejects [defaults.rig.imports.factory] in pack.toml. This migrates the Software Factory Intensive curriculum's active-lesson selection to the gc 1.2.1 shape (a [[rigs]] block carrying [rigs.imports.factory] in city.toml) as one coordinated change, so the curriculum's own linter, snapshots, walkthroughs, templates, and docs all move together. A partial migration breaks the repo's own checks, because the linter previously required the old pattern.

What changed

• test-harness/lesson-pack-lint.py: dropped the SFI112 rule that required the legacy pack.toml block. Added SFI104 (city config must select a lesson through a [[rigs]] factory import) and SFI114 (flags the legacy pack.toml block if it lingers). Re-pointed the SFI501/SFI502 doc checks at the [[rigs]] shape and fixed the SFI103 hint.
• my-factory/{pack,city}.toml.template: moved the factory import out of pack.toml and into the city.toml [[rigs]] block.
• test-harness/walkthrough-snapshots/L1/pack.toml: dropped the legacy block (the matching city.toml snapshot was already on the new shape).
• test-harness/walkthroughs/{L1,L2,L3,L4,C1}.sh: the generated city.toml now carries the [[rigs]] block, and the imperative gc --rig rig import add/remove factory step is replaced by a check that the import is declared in city.toml.
• Docs (~20) — the top-level README, my-factory/README.md, the activities/ and curriculum/ lab and capstone READMEs and PROMPTs, and the fired-up-pizza reference now teach editing city.toml instead of running gc --rig <rig> import.

How to verify

Run from the repo root:

python3 test-harness/lesson-pack-lint.py --lesson L2 --lesson L3 --lesson L4 --lesson C1
bash test-harness/migration-check.sh

The linter reports only the two pre-existing SFI100/SFI110 findings, which fire because the gitignored my-factory/{city,pack}.toml runtime copies do not exist on a fresh checkout. That matches the linter's behavior before this change. Copy the templates to real files (cp my-factory/pack.toml.template my-factory/pack.toml, same for city.toml) and the linter is fully clean, with zero findings. migration-check.sh passes every structural invariant.

Not verified: gc 1.2.1 runtime

Only gc 1.1.1 was available in the build environment, and 1.1.1 does not understand the [[rigs]] city.toml shape, so the walkthrough scripts' live gc register and gc rig add behavior under gc 1.2.1 was not exercised. The gc-independent checks above all pass. Confirm the walkthroughs against a real gc 1.2.1 before relying on them.

Open question for review

The [[rigs]] block declares a rig name, while gc rig add <dir> derives the rig name from the directory's basename. The generic templates and curriculum use name = "rig" (matching the committed L1/city.toml snapshot and the walkthrough's rig/ directory); the fired-up-pizza reference uses name = "fired-up-pizza". Whether the [[rigs]] name must match the gc rig add basename, and whether [[rigs]] needs a dir field pointing at the project, both need confirmation on gc 1.2.1.

Test plan

• lesson-pack-lint.py --lesson L2..C1 is clean except the pre-existing SFI100/SFI110
• migration-check.sh passes
• On gc 1.2.1, a walkthrough (e.g. test-harness/walkthroughs/L2.sh) registers the city and resolves the factory import from city.toml

---

Generated by the operator's software factory.
• City: factory-main · Agent: local-core.builder-2
• On behalf of: @austinborn